
Notion AI chats were exposed in a breach at AI vendor Fireworks
Notion has published a security advisory about a breach at Fireworks AI, a company it used to run some AI models. Notion says the incident touches less than 1% of Notion workspaces, and that its own systems were not breached.
Fireworks told Notion on November 6 that attackers may have reached customer data. The data held Notion AI chat transcripts that Notion used for Insights, a feature that shows admins how their teams use AI. For most affected workspaces, the chats carried no name or workspace label. Notion contacted affected customers on November 9.
Notion has turned off Insights while it investigates, and it no longer serves open-weight models through Fireworks. On November 11 it revoked Notion API keys it found in the exposed messages. On November 12 it reported other companies' API keys it found to the services that issued them.
Notion asks affected customers to treat the data as exposed. If they pasted passwords, API keys or other credentials into Notion AI in that period, they should change them.
What it means for your work
- For everyone: Keep passwords and keys out of AI chats. Use a password manager or a secrets tool.
- For affected teams: Read the data Notion gives you in your workspace, and rotate any credential you find in it.
- For IT and security leads: Ask each AI vendor which other companies process your chats and how long they keep them.
For more on keeping AI tools safe at work, see Google Cloud's new security tools for agents.